3x 30" Portrait 6970 CF Eyefinity vs 580 SLI Surround Showdown


I am on the notify list so I can pick up a 4th card. To think all those cards sold so fast and left me hanging with only three. :D Then again I don't have a MB that will support four yet unless I could get SLI hack to get it to work on my P67 board.

Gainward's 3GB GTX580 is available in the UK, at least:
http://www.scan.co.uk/products/3gb-...5-gpu-783mhz-shader-1566mhz-512-cores-dp-hdmi

Horrendous amount of money to pay for a graphics card though. I think it's fair to say these are only really useful when you have three of them, or two at the very least, and that makes them an extremely expensive option.

Vega: How long is 'super short'?

I would say the included 6 foot cords that came with my monitors pretty short. They are pretty beefy and include ferrite beads and I did not have signal problems with them. When I increased to after market 10-25 foot chords I had all sorts of signal problems.

Thanks for the Hard work Vega, truly appreciated by all of us :)

Two questions for you.

How much of the performance increase do you believe was from adding the third card in comparison to adding the extra memory?

Do you think the 3gb Vram will have an effect at 3600x1920? You inspired me so much my 3 ZR24w's are on the way :)

As you can see by a lot of the numbers, on the regular 580 the memory limit was really holding them down. Once you remove that limit and add a third card it really let everything work well. You can clearly see this in the Heaven 2.1 and A-10C numbers in which the 3GB cards scaled way past where they should have versus two regular 580's. Not just in FPS numbers but also smoothness.

You might be right on the border if 1.5GB will be enough or not. Depends on what games you play though and if you like AA or not. Some games use a ton more VRAM than others. If I was planning a Surround setup from scratch, I'd definitely spend the extra $100 per card for the 3GB.
 
vega;

first off, thanks for the efforts...truly herculean.

couple of questions:

1. what drivers did you settle on for the palits...it appears you've sorted them out; any portrait issues
with your current drivers?

2. did you test dual SLI with the palits? I am still thinking of getting a third zr30w and picking up
2 of the palit 580 3gb...if the scaling warrented, however, I may get a third...

3. did you ever hear back from EK re. waterblock for the palits?...I'm under water and don't want to switch
to loud GPUs without a solution.

thanks again,

rob
 
Ah right, 6 foot is ample for me, and I know to expect signal issues with longer cables even with HDMI and DVI (seen it happen often with both, even at 10-12 foot lengths). No issue for me, but perhaps it might be more of a problem with a wide eyefinity setup.
 
vega;

first off, thanks for the efforts...truly herculean.

couple of questions:

1. what drivers did you settle on for the palits...it appears you've sorted them out; any portrait issues
with your current drivers?

2. did you test dual SLI with the palits? I am still thinking of getting a third zr30w and picking up
2 of the palit 580 3gb...if the scaling warrented, however, I may get a third...

3. did you ever hear back from EK re. waterblock for the palits?...I'm under water and don't want to switch
to loud GPUs without a solution.

thanks again,

rob

1. I am using the very latest drivers, 266.58 with the caveat that I have to keep nVidia helper service turned off. I adjust settings with nVidia Inspector anyway.

2. I was going to do 2-way numbers but 3-way scaled so well I didn't bother. In the 3-way setup I was still getting 100% usage on all 3 GPU's with a 2600K at 4.6Ghz. I imagine once I get up to around 5.5Ghz on something like a 990x combined with a 4th 3GB 580, hopefully I won't be GPU limited and a little more balanced.

3. You can read this announcement: http://www.ekwaterblocks.com/index....tail,0&cntnt01articleid=63&cntnt01returnid=17
Their 580 line to include the 3GB version uses the same PCB as the 570 announced. I emailed them and they said they had a 3GB 580 inbound to make sure it also fits that. So I'd say it's a pretty safe bet. I'll be picking up four of the EK water blocks. The tricky part will be anti-condensation insulating them all with Dragonskin FX-pro for sub-zero liquid cooling. ;)

Ah right, 6 foot is ample for me, and I know to expect signal issues with longer cables even with HDMI and DVI (seen it happen often with both, even at 10-12 foot lengths). No issue for me, but perhaps it might be more of a problem with a wide eyefinity setup.

I use 25 foot high quality DVI-D cables since my computer is in a different room and have no issues with the signal at 2560x1600 per monitor.

Vega, how would you consider that the 3GB 580s scale in SLI?

Really excellent scaling. That is typical though at such high resolutions the GPU's get to flex their muscle without other parts of the system slowing them down.
 
Thanks for the benchmarks! I'm considering these cards as it's just as easy to exhaust memory in 3D 1920×1080 as it is at 2560×1600 2D. I've looked at it Afterburner and Heaven some time ago in 3D and 2xAA with everything else default pegs 1.5 GB even at only 5760x1080. My guess is that the benefit of 3 GB in 3D wouldn't be as great in 3D as I would think that there's simply more overhead in applying AA to two frames per scene versus one even though the frames are half the size.

These cards are still pretty rare so I'll have to look into when next gen cards arrive as well. Could be interesting.

Thanks Again!
 
Thanks for the benchmarks! I'm considering these cards as it's just as easy to exhaust memory in 3D 1920×1080 as it is at 2560×1600 2D. I've looked at it Afterburner and Heaven some time ago in 3D and 2xAA with everything else default pegs 1.5 GB even at only 5760x1080. My guess is that the benefit of 3 GB in 3D wouldn't be as great in 3D as I would think that there's simply more overhead in applying AA to two frames per scene versus one even though the frames are half the size.

These cards are still pretty rare so I'll have to look into when next gen cards arrive as well. Could be interesting.

Thanks Again!

Seems like the extra VRAM would serve you well then if your hitting the limit. Last I've read is an AMD/nVidia card refresh around this fall. Most things get delayed so I'd say closer to winter. Just from previous experience complete redesigns for new GPU's (not incremental like GTX480 to GTX580) take around 1.5 years.
 
I just finished doing the Vega monitor mod on my 2005FPW's. This took the distance between images from 1.75 inches to less than 1 inch and my bezel correction went from 180 pixels down to 90. I'm very pleased with the results, thanks for the idea Vega!



The bezel blindspots were pretty painful in game...


The actual image starts another 1/8th of an inch in from the bezel.


Relocated the OSD buttons to the bottom backside of the monitor.


Thin Bezels, I'm in Heaven!;)






I apologize in advance for all the pictures, I just wanted to show how big an improvement the setup you described in the OP makes.
 
Last edited:
^ - That's what I'm talking about. Bezel removed 3x portrait is pretty sweet isn't it.
 
Very sweet, I have a 37'' 3150x1680 display using ips monitors you could buy for $300-400 total.

I still want 30's at some point, likely once they get some 120hz models. The 1600 pixels of width on the center display would be much nicer for web content than my current setup.
 
Don't hold your breathe waiting on 120Hz 30" for many reasons. I think it will be many years from now and OLED might actually be on the scene replacing LCDs.
 
so Vega, in the end b/c of the Crossfire bridge limitation, you weren't able to add a 3rd 6970 to work @ all?

just wanted to make sure, as I was contemplating adding a 3rd
 
Don't hold your breathe waiting on 120Hz 30" for many reasons. I think it will be many years from now and OLED might actually be on the scene replacing LCDs.

Yeah, I'm still waiting on 1080P 120Hz IPS monitors, Mitsubishi released on in Japan last year but I've not heard much else. The bandwidth and GPU power to drive these guys in 3D which would be one of the drivers for these monitors is well beyond todays technology.
 
so Vega, in the end b/c of the Crossfire bridge limitation, you weren't able to add a 3rd 6970 to work @ all?

just wanted to make sure, as I was contemplating adding a 3rd

As you can see on the chart a few did work, but obviously I expect everything to work. Will you be getting a new MB? I don't think that MB in your sig can do 3x 6970's. If you get a 3rd, please post your results.

Yeah, I'm still waiting on 1080P 120Hz IPS monitors, Mitsubishi released on in Japan last year but I've not heard much else. The bandwidth and GPU power to drive these guys in 3D which would be one of the drivers for these monitors is well beyond todays technology.

Those Mitsubishi 120Hz IPS are just frame interpolation junk that they have put into TVs for years. They only accept a 60Hz input.
 
Those Mitsubishi 120Hz IPS are just frame interpolation junk that they have put into TVs for years. They only accept a 60Hz input.

Thanks, hadn't kept up with it as I hadn't heard anything much said about them. Oh well, the wait continues.
 
I just finished doing the Vega monitor mod on my 2005FPW's. This took the distance between images from 1.75 inches to less than 1 inch and my bezel correction went from 180 pixels down to 90. I'm very pleased with the results, thanks for the idea Vega!



The bezel blindspots were pretty painful in game...


The actual image starts another 1/8th of an inch in from the bezel.


Relocated the OSD buttons to the bottom backside of the monitor.


Thin Bezels, I'm in Heaven!;)






I apologize in advance for all the pictures, I just wanted to show how big an improvement the setup you described in the OP makes.

Why MUST YOU TORTURE ME!?!?!? Everytime I see something like this my mind says " Hey screw the 30 inch monitor and get two more LP2475w!" I swear I got some PC hardware related issue.
 
Quoting pictures. :mad:

Is an amazing setup, but personally, I'll wait until I can do it with three OLED 30"ers.
 
I have confirmed that Afterburner 2.1.0 Beta 7 can adjust voltages on these non-reference cards.
 
Cool, just let our grandchildren know how it looks when you get that set up. :p

Given what this one 30" cost me, even if the technology were available today, it'd be a fair few years before I could justify buying it :p
 
I have confirmed that Afterburner 2.1.0 Beta 7 can adjust voltages on these non-reference cards.
Score, you shooting for 1ghz with the waterblocks?

After spending some time with my portrait setup, I can't believe you were even able to run with 1.5gb of vram. I can't run 4x AA at 3330x1680 without bumping into my 1250mb of vram on the 470s.
 
Score, you shooting for 1ghz with the waterblocks?

After spending some time with my portrait setup, I can't believe you were even able to run with 1.5gb of vram. I can't run 4x AA at 3330x1680 without bumping into my 1250mb of vram on the 470s.

With sub-zero liquid I should be able to get 1000Mhz on all four 3GB 580s I think. Maybe get close to some records. :D

Liquid nitrogen! :D

I don't care for the temporary stuff much. ;) Shooting for a 24/7 sub-zero liquid cooling setup cooling CPU and 4x GPU's.
 
2. Download this program and hold onto it since God knows why Matrox even has it available for free (second link): http://www.matrox.com/graphics/en/support/gxm/software/?os=2&product=71

One of the utilities in there lets you carve out rectangles that windows will maximize to. So in the case of a spanned triple monitor, you carve it into thirds to make it act like each monitor is independent for 2D use. Nvidia lacks anything of the sort and AMD's Hydra Grid is a pale imitation.

Fail. Powerdesk does not allow RDP windows to maximize (i.e. showing the control bar, intercepting Windows keys) to a portion of the Eyefinity desktop. This is why I was interested in trying it. Actually, I disagree that Hydragrid is a pale imitation.... it's very similar and in fact has more functionality.
 
It"s "fail" because it doesn't work with some obscure feature that no one gives a fuck about in this context?

Why you would rely on an Eyefinity system for RDP no one knows.

Hydragrid is shit just like all the other terrible AMD software and UI implementation. Thankfully their drivers are at least functional...for Hydragrid to be useful AMD should have made it easier to maximize windows to grid instead of having to assign hotkeys or right click and make a window realize that Hydragrid exists.
 
It"s "fail" because it doesn't work with some obscure feature that no one gives a fuck about in this context?

"Fail" in that Matrox Powerdesk isn't actually any better than Hydragrid. Thanks for introducing profanity into the discussion. :rolleyes:

Why you would rely on an Eyefinity system for RDP no one knows.

Rely? I happen to want to run RDP on my system without switching profiles. You may not do any work on YOUR gaming system, but I do.

Hydragrid is shit just like all the other terrible AMD software and UI implementation. Thankfully their drivers are at least functional...for Hydragrid to be useful AMD should have made it easier to maximize windows to grid instead of having to assign hotkeys or right click and make a window realize that Hydragrid exists.

Sigh. It was a design choice to have to attach windows to the grid. I agree with it, it's more flexible than the Matrox solution.

CCC is shit because it was outsourced to Hyderabad.
 
Using what if I may ask?

Distilled water + watter wetter + Propylene Glycol. I know Propylene Glycol doesn't perform the best, gets slushy and what not super cold but its safe. I won't use anything toxic (I have pets) or gives off nasty fumes. Shooting for around -20/-30C should allow a pretty decent over clock on 580s and a CPU. The phase change is a 1.7HP compressor window A/C unit so I can dump all of the computer heat outside and not have my HVAC system have to deal with it.

Hopefully you have a 40A circuit, hoss

Instead of one massive power supply I will run two. This way they can easily be split over two separate standard 20A circuits.
 
Distilled water + watter wetter + Propylene Glycol. I know Propylene Glycol doesn't perform the best, gets slushy and what not super cold but its safe. I won't use anything toxic (I have pets) or gives off nasty fumes. Shooting for around -20/-30C should allow a pretty decent over clock on 580s and a CPU. The phase change is a 1.7HP compressor window A/C unit so I can dump all of the computer heat outside and not have my HVAC system have to deal with it.

Very [H]ard indeed! So what kind of blocks are you going to use on the 580s?
 
Distilled water + watter wetter + Propylene Glycol. I know Propylene Glycol doesn't perform the best, gets slushy and what not super cold but its safe. I won't use anything toxic (I have pets) or gives off nasty fumes. Shooting for around -20/-30C should allow a pretty decent over clock on 580s and a CPU. The phase change is a 1.7HP compressor window A/C unit so I can dump all of the computer heat outside and not have my HVAC system have to deal with it.



Instead of one massive power supply I will run two. This way they can easily be split over two separate standard 20A circuits.

Will you be submerging the cool side rad in the liquid solution? The was a guy on here who did a similar setup last sumer with just a cpu loop, I think he just used winter windshield washer fluid though.
 
Very [H]ard indeed! So what kind of blocks are you going to use on the 580s?

EK is making blocks for these new non-reference 580s.

Will you be submerging the cool side rad in the liquid solution? The was a guy on here who did a similar setup last sumer with just a cpu loop, I think he just used winter windshield washer fluid though.

The evaporator will be submerged yes. What I am doing is a bit unconventional. I am building the tank inside the window A/C unit surrounding the evaporator. It should be a nice and tidy solution, but a bit of work to fabricate! Windshield washer fluid has some pretty nasty stuff in it. Methanol, ethanol, ethylene glycol. All toxic stuff.
 
I don't know what sort of power your cooling system will use, but yeah, breaking the limit of a 20A line in the US is certainly possible. The DC load of the GPUs at 1Ghz should be around 1300-1400W, and the rest of the system's unlikely to draw more than 300W, so factoring in say 85% PSU efficiency with good units that's going to be about 2KW. Add maybe 110-120W each for the monitors, and at 110V you'd be using about 21.5A. Yikes.
Obviously with 1700W DC coming out, two PSUs is basically mandatory.
 
I don't know what sort of power your cooling system will use, but yeah, breaking the limit of a 20A line in the US is certainly possible. The DC load of the GPUs at 1Ghz should be around 1300-1400W, and the rest of the system's unlikely to draw more than 300W, so factoring in say 85% PSU efficiency with good units that's going to be about 2KW. Add maybe 110-120W each for the monitors, and at 110V you'd be using about 21.5A. Yikes.
Obviously with 1700W DC coming out, two PSUs is basically mandatory.

Since the Antec TPQ 1200 isn't the most efficient PSU out there, I was thinking about selling it off for 2x 1200w Highpower HP-1200-G14S(C)-PT which are Platinum certified. I have an email in with them to see if they are for sale yet. I know it will likely take me about 5 years to regain the Ebay loss on the TPQ1200 but I do like energy efficiency, clearly. My computer will just sip electricity. :D At least my local power is nuclear!
 
Why not go for a pair of Corsair AX 1200s? They're phenomenally efficient, 89.5% at 1200W DC on UK voltage, and still pushing 86% when over-run all the way to 1520W DC.
 
Why not go for a pair of Corsair AX 1200s? They're phenomenally efficient, 89.5% at 1200W DC on UK voltage, and still pushing 86% when over-run all the way to 1520W DC.

If the High Power platinums aren't for sale or are stupidly expensive, I'll most likely get 2x Antec HCP-1200's. The Corsair is nice but the Antec edges it out ever so slightly in reviews. Plus I prefer multiple rail over current protection versus single rail protection (which at over 100 AMPs isn't much protection.

Still debating whether to have one +12v rail dedicated to each GPU with the +8/+6 pin PCI-e connector shared on that single rail or split each of the +8/+6 over multiple +12v rails. I've read both pros and cons of both. Not sure how 2x PSUs will work into that though.
 
Yeah, 100 amp OCP is pretty dumb. It does the exact opposite of what a high-end PSU should do, which is protect your components.
 
Getting rid of this Sandy Bridge stuff. Have a 4th 3GB 580 inbound and a lightly used GA-X58-UD9 on the way for $380 bucks. I think its a pretty sweet deal for 4-way SLI board with Sata 6GB/USB 3.0
 
Back
Top